I just posted these two installments to the Digital Users Group
mailing list, but I thought PDMLers might be interested.
G
---- #1 ----
Before I sold the DA16-45lens, I spent an hour or so shooting a bunch
of test exposures with it, the FA20-35, the FA28-105, the A24, and
the F35-70. I had only a 512M card in the camera so I switched to
JPEG *** and did a sequence with each lens at the focal lengths 20,
24, 28, 35 and 45mm, or as close to those I could estimate with the
zoom rings, with the camera mounted on a tripod looking at the same
subject.
Each series of exposures included setting aperture to wide open and
going from there in the standard aperture sequence to f/22. All at
ISO 200, Av mode. (BTW: while there is some variation in exposure
going aperture to aperture in all of the series, it isn't terribly
great and it is hardly any different or more variable for the DA16-45
than for any of the others. Exposure for all the lenses and all the
focal lengths and apertures is remarkably consistent...)
The results of that testing I'll put up on a web page soon.
But what I saw in my JPEGs was that my DS has developed a hot pixel
slightly to left of center, which becomes evident at approximately
1/8 second or longer exposures. I was curious about this as I'd not
seen it before. So this morning I shot a few JPEG *** frames at 1,
1/2, 1/4, 1/8, 1/15, and 1/30 second, just setting the camera on
manual focus and exposure and snapping away at a dim corner of my
office, with noise reduction off. Sure enough, there's my hot pixel.
In a curious moment, I switched the camera to RAW format capture and
did the same sequence. Result: the hot pixel is gone, both from the
embedded JPEG previews and from the RAW converted, [EMAIL PROTECTED]
image, converted with Camera Raw set to As Shot, all Auto processing
and noise reduction turned off.
I'm not exactly sure what to make of this... But at least if I am
storing photos in RAW format, I guess I don't have to worry about
*that* hot pixel!!
---- #2 ----
Follow up:
I made another test sequence, all in "natural color mode" set on,
manual exposure, 1 second exposure. These are the results:
sRGB - NR on :: no hot pixel
sRGB - NR off :: hot pixel
ARGB - NR on :: no hot pixel
ARGB - NR off :: hot pixel (actually, two!)
RAW - NR on :: no hot pixel
RAW - NR off :: no hot pixel
I'm using Adobe Camera Raw v3.1 with Photoshop CS2, and I turned off
all color noise reduction and luminance smoothing in the Details
slider. I don't know what other hot pixel removal it might be doing
as a standard part of processing beyond that. But the in-camera JPEG
previews do not show the hot pixel in the RAW files while they do in
the in-camera JPEG *** files.
Then someone on DPReview suggested:
WeberPoint:
> It might be interesting to convert the RAW files with RSE, or one
> of the other free converters, to see if the behavior is the same.
Good suggestion. RSE isn't available on Mac OS X, but I have both
Pentax Lab v2.1 and Vuescan v8.2.10.
I took the two PEF files (one with NR turned on and one with NR
turned off) and processed them with both Pentax Lab and Vuescan.
Pentax Lap produced [EMAIL PROTECTED] TIFF files without any hot pixels
in evidence, and they looked pretty similar.
Vuescan, on the other hand, with all filtering and color adjustment
turned off, produced [EMAIL PROTECTED] RGB TIFF files that looked
drastically different from all the previous ones. The PEF made with
NR turned on shows no hot pixels, but there are quite a few artifacts
which obviously came from the NR dark frame subtraction ... remember,
this is an arbitrary 1second exposure of a dark corner of the room,
not focussed, at ISO 200, f/8, 1 second. I expected it to be
underexposed and the PEF-NR file shows pretty much what I expected in
that situation.
The PEF file with NR turned off, however, produced a surprising TIFF
file: Vuescan's RGB rendering shows smooth, clean blacks with very
little blotchiness or color noise, smooth transitions, and three hot
pixels, including the elusive one that I have been looking at. The
other two are "less hot" ... the one I've been looking for shows up
at 1/8 second, faintly ... at 8x that exposure, the other two seem to
surface.
My conclusions from all this testing and investigation:
1) I do have at least one, if not three, hot pixels.
2) Normal capture into RAW format with NR off and processing
with ACR 3.1 (or PL 2.1) eliminates them. Thus ACR and PL
must be doing some kind of hot pixel removal.
3) Capture in-camera into JPEG with NR on eliminates them.
4) NR on has the possibility of adding noise in undexposure
circumstances.
5) Vuescan likely does a simpler, purer job of RAW conversion
processing than ACR or PL. (I know that Vuescan's RAW
conversion code is from the dcraw source code library,
so dcraw is probably doing the same, simple RAW conversion.)
I have yet to decide whether I consider the hot pixels important
enough to send the camera in for service, given that my usual RAW
image processing workflow eliminates them. I might hold off until I
purchase my planned second body as otherwise I'll be forced to pick
up my 10D again in the absence of the DS...
Heaven forfend. ]'-)
Godfrey
